例如，我有一个张量为public class Startup : FunctionsStartup
{
public override void Configure(IFunctionsHostBuilder builder)
{
// Get the path to the folder that has appsettings.json and other files.
// Note that there is a better way to get this path: ExecutionContext.FunctionAppDirectory when running inside a function. But we don't have access to the ExecutionContext here.
// Functions team should improve this in future. It will hopefully expose FunctionAppDirectory through some other way or env variable.
string basePath = IsDevelopmentEnvironment() ?
Environment.GetEnvironmentVariable("AzureWebJobsScriptRoot") :
$"{Environment.GetEnvironmentVariable("HOME")}\\site\\wwwroot";
var config = new ConfigurationBuilder()
.SetBasePath(basePath)
.AddJsonFile("appsettings.json", optional: false, reloadOnChange: false) // common settings go here.
.AddJsonFile($"appsettings.{Environment.GetEnvironmentVariable("AZURE_FUNCTIONS_ENVIRONMENT")}.json", optional: false, reloadOnChange: false) // environment specific settings go here
.AddJsonFile("local.settings.json", optional: true, reloadOnChange: false) // secrets go here. This file is excluded from source control.
.AddEnvironmentVariables()
.Build();
builder.Services.AddSingleton<IConfiguration>(config);
}
public bool IsDevelopmentEnvironment()
{
return "Development".Equals(Environment.GetEnvironmentVariable("AZURE_FUNCTIONS_ENVIRONMENT"), StringComparison.OrdinalIgnoreCase);
}
}
的张量，并且我想将(None, 2, 100, 100, 1024)分为2和1，这样我就有2个张量为4 {{ 1}}。如何使用Keras Tensorflow做到这一点？
谢谢。
答案 0 :(得分：1)
使用tf.split()：
import tensorflow as tf
tensor = tf.placeholder(tf.float32, (None, 2, 100, 100, 1024))
splitted = [tf.squeeze(t, axis=1) for t in tf.split(tensor, 2, axis=1)]
print(splitted[0].get_shape().as_list(), splitted[1].get_shape().as_list())
# [None, 100, 100, 1024] [None, 100, 100, 1024]
要串联起来：
# manipulate here ...
splitted = [t[:, None, ...] for t in splitted]
res = tf.concat(splitted, axis=1)
print(res.get_shape().as_list()) # [None, 2, 100, 100, 1024]